[Top] [All Lists]

Re: [PATCH 2/4] renameat2 syscall: check plain rename

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Subject: Re: [PATCH 2/4] renameat2 syscall: check plain rename
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Thu, 17 Apr 2014 09:03:15 +1000
Cc: Miklos Szeredi <miklos@xxxxxxxxxx>, Miklos Szeredi <mszeredi@xxxxxxx>, xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <20140416141825.GA32449@xxxxxxxxxxxxx>
References: <1397231517-4641-1-git-send-email-miklos@xxxxxxxxxx> <1397231517-4641-2-git-send-email-miklos@xxxxxxxxxx> <20140414023020.GJ27694@dastard> <20140416141825.GA32449@xxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Wed, Apr 16, 2014 at 07:18:25AM -0700, Christoph Hellwig wrote:
> On Mon, Apr 14, 2014 at 12:30:20PM +1000, Dave Chinner wrote:
> > IOWs, XFS is returning EEXIST rather than ENOTEMPTY for several of
> > these rename tests. The rename man page says this about the errors:
> > 
> >        ENOTEMPTY or EEXIST
> >               newpath is a nonempty directory, that is, contains
> >           entries other than "." and "..".
> > 
> > Which implies that both errors are valid and so the test should pass
> > in either case. Can you send a patch to handle these
> > different-but-valid error returns?
> I would much prefer if all Linux filesystems behaved uniformly here.
> While EEXIST sounds much more logical to me in this case I suspect most
> other filesystems have copy & pasted from ext2, and we should switch to
> ENOTEMPTY as well.

Send patches, please.

But from an xfstests perspective, we still have to support kernels
and filesystems that return EEXIST for this function, as it is
valid for them to do so....


Dave Chinner

<Prev in Thread] Current Thread [Next in Thread>